Bert pre-train 代码

目录

Pre-train主体代码结构

Initial Config

Build model

Masked LM预测

Next Sentence 预测

Bert主体

Input_fn

Initial Estimator

Train

---

Pre-train主体代码结构

def main(_):
  tf.logging.set_verbosity(tf.logging.INFO)

  if not FLAGS.do_train and not FLAGS.do_eval:
    raise ValueError("At least one of `do_train` or `do_eval` must be True.")

  bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)

  tf.gfile.MakeDirs(FLAGS.output_dir)

  input_files = []
  for input_pattern in FLAGS.input_file.split(","):
    input_files.extend(tf.gfile.Glob(input_pattern))

  tf.logging.info("*** Input Files ***")
  for input_file in input_files:
    tf.logging.info("  %s" % input_file)

  tpu_cluster_resolver = None
  if FLAGS.use_tpu and FLAGS.tpu_name:
    tpu_cluster_resolver = tf.contrib.cluster_resolver.TPUClusterResolver(
        FLAGS.tpu_name, zone=FLAGS.tpu_zone, project=FLAGS.gcp_project)

  is_per_host = tf.contrib.tpu.InputPipelineConfig.PER_HOST_V2
  run_config = tf.contrib.tpu.RunConfig(
      cluster=tpu_cluster_resolver,
      master=FLAGS.master,
      model_dir=FLAGS.output_dir,
      save_checkpoints_steps=FLAGS.save_checkpoints_steps,
      tpu_config=tf.contrib.tpu.TPUConfig(
          iterations_per_loop=FLAGS.iterations_per_loop,
          num_shards=FLAGS.num_tpu_cores,
          per_host_input_for_training=is_per_host))

  model_fn = model_fn_builder(
      bert_config=bert_config,
      init_checkpoint=FLAGS.init_checkpoint,
      learning_rate=FLAGS.learning_rate,
      num_train_steps=FLAGS.num_train_steps,
      num_warmup_steps=FLAGS.num_warmup_steps,
      use_tpu=FLAGS.use_tpu,
      use_one_hot_embeddings=FLAGS.use_tpu)

  # If TPU is not available, this will fall back to normal Estimator on CPU
  # or GPU.
  estimator = tf.contrib.tpu.TPUEstimator(
      use_tpu=FLAGS.use_tpu,
      model_fn=model_fn,
      config=run_config,
      train_batch_size=FLAGS.train_batch_size,
      eval_batch_size=FLAGS.eval_batch_size)

  if FLAGS.do_train:
    tf.logging.info("***** Running training *****")
    tf.logging.info("  Batch size = %d", FLAGS.train_batch_size)
    train_input_fn = input_fn_builder(
        input_files=input_files,
        max_seq_length=FLAGS.max_seq_length,
        max_predictions_per_seq=FLAGS.max_predictions_per_seq,
        is_training=True)
    estimator.train(input_fn=train_input_fn, max_steps=FLAGS.num_train_steps)

  if FLAGS.do_eval:
    tf.logging.info("***** Running evaluation *****")
    tf.logging.info("  Batch size = %d", FLAGS.eval_batch_size)

    eval_input_fn = input_fn_builder(
        input_files=input_files,
        max_seq_length=FLAGS.max_seq_length,
        max_predictions_per_seq=FLAGS.max_predictions_per_seq,
        is_training=False)

    result = estimator.evaluate(
        input_fn=eval_input_fn, steps=FLAGS.max_eval_steps)

    output_eval_file = os.path.join(FLAGS.output_dir, "eval_results.txt")
    with tf.gfile.GFile(output_eval_file, "w") as writer:
      tf.logging.info("***** Eval results *****")
      for key in sorted(result.keys()):
        tf.logging.info("  %s = %s", key, str(result[key]))
        writer.write("%s = %s\n" % (key, str(result[key])))

Initial Config

Bert model config, session config, distribution strategy, 将前面这些config传入给Run_config

# config code
bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)

# Creates session config. allow_soft_placement = True, is required for
# multi-GPU and is not harmful for other modes.
session_config = tf.ConfigProto(
    inter_op_parallelism_threads=FLAGS.inter_op_parallelism_threads,
    intra_op_parallelism_threads=FLAGS.intra_op_parallelism_threads,
    allow_soft_placement=True)

distribution_strategy = distribution_utils.get_distribution_strategy(
    get_num_gpus(FLAGS), FLAGS.all_reduce_alg)

# Creates a `RunConfig` that checkpoints every 24 hours which essentially
# results in checkpoints determined only by `epochs_between_evals`.
# for tmp test save_checkpoints_secs = 60*60, each hour
run_config = tf.estimator.RunConfig(
    train_distribute=distribution_strategy,
    session_config=session_config,
    model_dir=FLAGS.output_dir,
    save_checkpoints_steps=FLAGS.save_checkpoints_steps,
)

Build model

model_fn = model_fn_builder()

获取数据内容，传入到Bert主体(12层Transformer)中。对下一句预测任务取出模型的[CLS]结果。对遮蔽词预测任务取出模型的最后结果。然后分别计算loss值，最后将loss值相加。

• Masked LM预测

获取BERT模型的最后一层序列的encoder tensors，输出遮蔽词预测任务的loss和概率矩阵。

# MLM
def get_masked_lm_output(bert_config, input_tensor, output_weights, positions,
                         label_ids, label_weights):
  """Get loss and log probs for the masked LM."""
  input_tensor = gather_indexes(input_tensor, positions)

  with tf.variable_scope("cls/predictions"):
    # We apply one more non-linear transformation before the output layer.
    # This matrix is not used after pre-training.
    with tf.variable_scope("transform"):
      input_tensor = tf.layers.dense(
          input_tensor,
          units=bert_config.hidden_size,
          activation=modeling.get_activation(bert_config.hidden_act),
          kernel_initializer=modeling.create_initializer(
              bert_config.initializer_range))
      input_tensor = modeling.layer_norm(input_tensor)

    # The output weights are the same as the input embeddings, but there is
    # an output-only bias for each token.
    output_bias = tf.get_variable(
        "output_bias",
        shape=[bert_config.vocab_size],
        initializer=tf.zeros_initializer())
    logits = tf.matmul(input_tensor, output_weights, transpose_b=True)
    logits = tf.nn.bias_add(logits, output_bias)
    log_probs = tf.nn.log_softmax(logits, axis=-1)

    label_ids = tf.reshape(label_ids, [-1])
    label_weights = tf.reshape(label_weights, [-1])

    one_hot_labels = tf.one_hot(
        label_ids, depth=bert_config.vocab_size, dtype=tf.float32)

    # The `positions` tensor might be zero-padded (if the sequence is too
    # short to have the maximum number of predictions). The `label_weights`
    # tensor has a value of 1.0 for every real prediction and 0.0 for the
    # padding predictions.
    per_example_loss = -tf.reduce_sum(log_probs * one_hot_labels, axis=[-1])
    numerator = tf.reduce_sum(label_weights * per_example_loss)
    denominator = tf.reduce_sum(label_weights) + 1e-5
    loss = numerator / denominator

  return (loss, per_example_loss, log_probs)

 

• Next Sentence 预测

获取cls位的数据，建立一层全连接层，再接一层softmax分类层

# Next Sentence
def get_masked_lm_output(bert_config, input_tensor, output_weights, positions,
                         label_ids, label_weights):
  """Get loss and log probs for the masked LM."""
  input_tensor = gather_indexes(input_tensor, positions)

  with tf.variable_scope("cls/predictions"):
    # We apply one more non-linear transformation before the output layer.
    # This matrix is not used after pre-training.
    with tf.variable_scope("transform"):
      input_tensor = tf.layers.dense(
          input_tensor,
          units=bert_config.hidden_size,
          activation=modeling.get_activation(bert_config.hidden_act),
          kernel_initializer=modeling.create_initializer(
              bert_config.initializer_range))
      input_tensor = modeling.layer_norm(input_tensor)

    # The output weights are the same as the input embeddings, but there is
    # an output-only bias for each token.
    output_bias = tf.get_variable(
        "output_bias",
        shape=[bert_config.vocab_size],
        initializer=tf.zeros_initializer())
    logits = tf.matmul(input_tensor, output_weights, transpose_b=True)
    logits = tf.nn.bias_add(logits, output_bias)
    log_probs = tf.nn.log_softmax(logits, axis=-1)

    label_ids = tf.reshape(label_ids, [-1])
    label_weights = tf.reshape(label_weights, [-1])

    one_hot_labels = tf.one_hot(
        label_ids, depth=bert_config.vocab_size, dtype=tf.float32)

    # The `positions` tensor might be zero-padded (if the sequence is too
    # short to have the maximum number of predictions). The `label_weights`
    # tensor has a value of 1.0 for every real prediction and 0.0 for the
    # padding predictions.
    per_example_loss = -tf.reduce_sum(log_probs * one_hot_labels, axis=[-1])
    numerator = tf.reduce_sum(label_weights * per_example_loss)
    denominator = tf.reduce_sum(label_weights) + 1e-5
    loss = numerator / denominator

  return (loss, per_example_loss, log_probs)

Bert主体

# Bert Model
with tf.variable_scope("bert", scope):
  with tf.variable_scope("embeddings"):
    # Perform embedding lookup on the word ids.
    (self.embedding_output, self.embedding_table) = embedding_lookup(
        input_ids=input_ids,
        vocab_size=config.vocab_size,
        embedding_size=config.hidden_size,
        initializer_range=config.initializer_range,
        word_embedding_name="word_embeddings",
        use_one_hot_embeddings=use_one_hot_embeddings)

    # Add positional embeddings and token type embeddings, then layer
    # normalize and perform dropout.
    self.embedding_output = embedding_postprocessor(
        input_tensor=self.embedding_output,
        use_token_type=True,
        token_type_ids=token_type_ids,
        token_type_vocab_size=config.type_vocab_size,
        token_type_embedding_name="token_type_embeddings",
        use_position_embeddings=True,
        position_embedding_name="position_embeddings",
        initializer_range=config.initializer_range,
        max_position_embeddings=config.max_position_embeddings,
        dropout_prob=config.hidden_dropout_prob)

  with tf.variable_scope("encoder"):
    # This converts a 2D mask of shape [batch_size, seq_length] to a 3D
    # mask of shape [batch_size, seq_length, seq_length] which is used
    # for the attention scores.
    attention_mask = create_attention_mask_from_input_mask(
        input_ids, input_mask)

    # Run the stacked transformer.
    # `sequence_output` shape = [batch_size, seq_length, hidden_size].
    self.all_encoder_layers = transformer_model(
        input_tensor=self.embedding_output,
        attention_mask=attention_mask,
        hidden_size=config.hidden_size,
        num_hidden_layers=config.num_hidden_layers,
        num_attention_heads=config.num_attention_heads,
        intermediate_size=config.intermediate_size,
        intermediate_act_fn=get_activation(config.hidden_act),
        hidden_dropout_prob=config.hidden_dropout_prob,
        attention_probs_dropout_prob=config.attention_probs_dropout_prob,
        initializer_range=config.initializer_range,
        do_return_all_layers=True)

  self.sequence_output = self.all_encoder_layers[-1]
  # The "pooler" converts the encoded sequence tensor of shape
  # [batch_size, seq_length, hidden_size] to a tensor of shape
  # [batch_size, hidden_size]. This is necessary for segment-level
  # (or segment-pair-level) classification tasks where we need a fixed
  # dimensional representation of the segment.
  with tf.variable_scope("pooler"):
    # We "pool" the model by simply taking the hidden state corresponding
    # to the first token. We assume that this has been pre-trained
    first_token_tensor = tf.squeeze(self.sequence_output[:, 0:1, :], axis=1)
    self.pooled_output = tf.layers.dense(
        first_token_tensor,
        config.hidden_size,
        activation=tf.tanh,
        kernel_initializer=create_initializer(config.initializer_range))

Input_fn

数据规范化

Initial Estimator

新建Estimator:

Train

Train: Estimator.Train

Optimizer: 将loss值输入给优化方法，优化并update weight。Training flow如下图

# train op
def create_optimizer(loss, init_lr, num_train_steps, num_warmup_steps, use_tpu):
  """Creates an optimizer training op."""
  global_step = tf.train.get_or_create_global_step()

  learning_rate = tf.constant(value=init_lr, shape=[], dtype=tf.float32)

  # Implements linear decay of the learning rate.
  learning_rate = tf.train.polynomial_decay(
      learning_rate,
      global_step,
      num_train_steps,
      end_learning_rate=0.0,
      power=1.0,
      cycle=False)

  # Implements linear warmup. I.e., if global_step < num_warmup_steps, the
  # learning rate will be `global_step/num_warmup_steps * init_lr`.
  if num_warmup_steps:
    global_steps_int = tf.cast(global_step, tf.int32)
    warmup_steps_int = tf.constant(num_warmup_steps, dtype=tf.int32)

    global_steps_float = tf.cast(global_steps_int, tf.float32)
    warmup_steps_float = tf.cast(warmup_steps_int, tf.float32)

    warmup_percent_done = global_steps_float / warmup_steps_float
    warmup_learning_rate = init_lr * warmup_percent_done

    is_warmup = tf.cast(global_steps_int < warmup_steps_int, tf.float32)
    learning_rate = (
        (1.0 - is_warmup) * learning_rate + is_warmup * warmup_learning_rate)

  # create tensor name for logging
  tf.identity(learning_rate, name='learning_rate')
  tf.summary.scalar('learning_rate', learning_rate)

  # It is recommended that you use this optimizer for fine tuning, since this
  # is how the model was trained (note that the Adam m/v variables are NOT
  # loaded from init_checkpoint.)
  optimizer = AdamWeightDecayOptimizer(
      learning_rate=learning_rate,
      weight_decay_rate=0.01,
      beta_1=0.9,
      beta_2=0.999,
      epsilon=1e-6,
      exclude_from_weight_decay=["LayerNorm", "layer_norm", "bias"])

  if use_tpu:
    optimizer = tf.contrib.tpu.CrossShardOptimizer(optimizer)

  tvars = tf.trainable_variables()
  grads = tf.gradients(loss, tvars)

  # This is how the model was pre-trained.
  (grads, _) = tf.clip_by_global_norm(grads, clip_norm=1.0)

  train_op = optimizer.apply_gradients(
      zip(grads, tvars), global_step=global_step)

  new_global_step = global_step + 1
  train_op = tf.group(train_op, [global_step.assign(new_global_step)])
  return train_op